Re: [PATCH 2/6] locking: Introduce range reader/writer lock

[Davidlohr Bueso]

On Tue, 18 Apr 2017, Laurent Dufour wrote:

> On 06/04/2017 10:46, Davidlohr Bueso wrote:
> > +__range_read_lock_common(struct range_rwlock_tree *tree,
> > +			 struct range_rwlock *lock, long state)
> > +{
> > +	struct interval_tree_node *node;
> > +	unsigned long flags;
> > +
> > +	spin_lock_irqsave(&tree->lock, flags);
> > +	range_lock_set_reader(lock);
> > +
> > +	if (!__range_intersects_intree(tree, lock))
> > +		goto insert;
> > +
> > +	range_interval_tree_foreach(node, &tree->root,
> > +				    lock->node.start, lock->node.last) {
> > +		struct range_rwlock *blocked_lock;
> > +		blocked_lock = range_entry(node, struct range_rwlock, node);
> > +
> > +		if (!range_lock_is_reader(blocked_lock))
> > +			lock->blocking_ranges++;
> > +	}
> > +insert:
> > +	__range_tree_insert(tree, lock);
> > +
> > +	lock->waiter = current;
>
> Hi Davidlohr,
>
> Setting lock->waiter after calling range_lock_set_reader() is resetting
> the reader flag. Moving the call to range_lock_set_reader() here fixes that.

Yeah, I moved the set_reader() call to after setting lock->waiter.

> > +	spin_unlock_irqrestore(&tree->lock, flags);
> > +
> > +	return wait_for_ranges(tree, lock, state);
> > +}

[...]

> > +int range_read_trylock(struct range_rwlock_tree *tree, struct range_rwlock *lock)
> > +{
> > +	int ret = true;
> > +	unsigned long flags;
> > +	struct interval_tree_node *node;
> > +
> > +	spin_lock_irqsave(&tree->lock, flags);
> > +
> > +	if (!__range_intersects_intree(tree, lock))
> > +		goto insert;
> > +
> > +	/*
> > +	 * We have overlapping ranges in the tree, ensure that we can
> > +	 * in fact share the lock.
> > +	 */
> > +	range_interval_tree_foreach(node, &tree->root,
> > +				    lock->node.start, lock->node.last) {
> > +		struct range_rwlock *blocked_lock;
> > +		blocked_lock = range_entry(node, struct range_rwlock, node);
> > +
> > +		if (!range_lock_is_reader(blocked_lock)) {
> > +			ret = false;
> > +			goto unlock;
> > +		}
> > +	}
> > +insert:
> > +	range_lock_set_reader(lock);
>
> Here, the lock->waiter field should have been set to current before
> calling range_lock_set_reader()

But this is a trylock attempt, there is no waiting going on.

Thanks,
Davidlohr